On some combinatorial problems in metric spaces of bounded doubling dimension


 Elaine Richardson
 11 months ago
 Views:
Transcription
1 On some combinatorial problems in metric spaces of bounded doubling dimension Michiel Smid February 5, 2010 Abstract A metric space has doubling dimension d if for every ρ > 0, every ball of radius ρ can be covered by at most 2 d balls of radius ρ/2. This generalizes the Euclidean dimension, because the doubling dimension of Euclidean space R d is proportional to d. The following results are shown, for any d 1 and any metric space of size n and doubling dimension d: First, the maximum number of diametral pairs is Θ(n 2 ). Second, if d = 1, the maximum possible weights of the minimum spanning tree and the allnearest neighbors graph are Θ(R log n) and Θ(R), respectively, where R is the minimum radius of any ball containing all elements of the metric space. Finally, if d > 1, the maximum possible weights of both the minimum spanning tree and the allnearest neighbors graph are Θ(Rn 1 1/d ). These results show that, for 1 d 3, metric spaces of doubling dimension d behave differently than their Euclidean counterparts. 1 Introduction Let S be a set of n points in R d, where d 1 is a constant. Let G be a graph with vertex set S, in which the weight of any edge (p, q) is defined to be the Euclidean distance pq between p and q. The weight wt(g) of G is defined to be the sum of the weights of the edges in G. School of Computer Science, Carleton University, Ottawa, Canada. This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC). 1
2 We are interested in the maximum values (over all sets of n points) of the number of diametral pairs, the weight of the minimum spanning tree, and the weight of the allnearest neighbors graph. Diametral pairs: The diameter of S is defined to be the maximum distance between any pair of points in S. Two points p and q in S form a diametral pair, if pq is equal to the diameter of S. Let DP(S) be the number of diametral pairs in the set S. Obviously, DP(S) = 1 if d = 1. If d = 2, the maximum possible value of DP(S) is n, whereas if d = 3, the maximum possible value is 2n 2. If d 4, the value of DP(S) can be as large as Θ(n 2 ). For proofs of these claims, see Chapter 13 in Pach and Agarwal [5]. Minimum spanning tree: If S is contained in the ddimensional cube [0, L] d, then the weight wt(mst (S)) of its minimum spanning tree MST (S) is O(n 1 1/d L). This result was first shown by Few [2]. Subsequently, Smith [7], and Steele and Snyder [8] gave tighter analyses in terms of the constant factor in the BigOh bound. For the set S consisting of the n vertices on an n 1/d... n 1/d grid with cells having sides of length n 1/d L, we have wt(mst (S)) = Ω(n 1 1/d L). Thus, the above upper bound on wt(mst (S)) is tight. Allnearest neighbors graph: For any point p in S, let NN S (p) denote a nearest neighbor of p in S. That is, NN S (p) is a point q in S \ {p} for which the Euclidean distance pq is minimum. The allnearest neighbors graph ANN (S) is defined to be the directed graph with vertex set S and edge set {(p, NN S (p)) : p S}. Assume again that the point set S is contained in the cube [0, L] d. Since ANN (S) is a subgraph of MST (S), we have wt(ann (S)) = O(n 1 1/d L). The grid example given above shows that this upper bound is tight. All results mentioned above are valid in the ddimensional Euclidean space. A natural question to ask is whether or not similar results hold in an arbitrary metric space. Let S be a finite set of elements, called points, that has a distance function associated with it. This function assigns to any two points p and q in S a real number pq, which is called the distance between p and q. The set S is called a metric space, if for all points p, q, and r in S, 2
3 1. pp = 0, 2. if p q, then pq > 0, 3. pq = qp, and 4. pq pr + rq. The fourth property is called the triangle inequality. If c is a point in S and R > 0 is a real number, then the ball with center c and radius R is defined to be the set {p S : cp R}. The radius of a subset S of S is defined to be the minimum radius of any ball that contains S. Observe that the center of this ball is not necessarily an element of S. For example, the set S of n points, in which each pair of distinct points has distance R, is a metric space of radius R. Obviously, we have DP(S) = ( n 2), wt(mst (S)) = (n 1)R, and wt(ann (S)) = nr. Moreover, these are the largest possible values in terms of n and R. Thus, in order to get nontrivial results, we have to consider restricted classes of metric spaces. In this paper, we consider metric spaces of bounded doubling dimension, which were introduced by Assouad [1]; see also the book by Heinonen [4]. Definition 1 Let S be a finite metric space and let λ be the smallest integer such that the following is true: For each real number ρ > 0, every ball in S of radius ρ can be covered by at most λ balls of radius ρ/2. The doubling dimension of the metric space S is defined to be log λ. Observe that every metric space of n points has doubling dimension at most log n. Furthermore, it is not difficult to prove that the doubling dimension of the Euclidean metric in R d is proportional to d. In this paper, we prove the following results, for any metric space S consisting of n points and having doubling dimension d: 1. If d 1, then DP(S) is Θ(n 2 ) in the worst case. Thus, the maximum possible value of DP(S) is very different from the corresponding value in Euclidean space R d for 1 d 3. In fact, we show that, for d = 1, the maximum possible value of DP(S) is n 2 /4. For d 4, the bound coincides with the corresponding bound for R d. 3
4 2. If d = 1, then wt(mst (S)) is Θ(R log n) in the worst case, whereas wt(ann (S)) is Θ(R) in the worst case. Thus, metric spaces of doubling dimension 1 behave differently than their onedimensional Euclidean counterparts. 3. If d > 1, then both wt(mst (S)) and wt(ann (S)) are Θ(Rn 1 1/d ) in the worst case. Thus, these bounds coincide with the corresponding bounds for Euclidean space R d. Related work: In recent years, metric spaces of bounded doubling dimension have received a lot of attention in the algorithms literature; see HarPeled and Mendel [3] and the references given in that paper. In [3], it was already observed that these spaces are different from their Euclidean counterparts: In Euclidean space R d, the allnearest neighbors graph can be computed in O(n log n) time, whereas the worstcase running time becomes Θ(n 2 ) even if the doubling dimension is equal to 1. Smid [6] defined the following metric space, based on an earlier example in [3]: Let S = {p 1, p 2,..., p n } and let 0 < ɛ < 1 be a real number. For each i and j with 1 i j n, we define 0 if i = j, p i p j = p j p i = 4 j if i = 1 and j > 1, 4 j + ɛ otherwise. Then S is a metric space of doubling dimension 1 and radius R = 4 n + ɛ. We have NN S (p 1 ) = p 2 and, for each j with 2 j n, NN S (p j ) = p 1. Thus, both ANN (S) and MST (S) are equal to the stargraph consisting of all edges (p 1, p j ), 2 j n. (In contrast, it is well known that in Euclidean space R d, the maximum degrees of both these graphs are upper bounded by a function that depends only on d.) Both wt(ann (S)) and wt(mst (S)) are close to 4R/3. The rest of this paper is organized as follows. In Section 2, we prove the upper bounds on DP(S), wt(ann (S)), and wt(mst (S)), for any finite metric space of doubling dimension d. The proofs will be by induction on the number of points in S. Some care has to be taken, because the doubling dimension d of a subset of S may be larger than d. We show that d is always at most 2d. 4
5 In Section 3, we prove that the upper bounds in Section 2 are tight. The lower bounds for DP(S), wt(ann (S)), and wt(mst (S)) are in fact obtained by one single class of metric spaces. We conclude in Section 4 by presenting some more results that show that metric spaces of doubling dimension 1 are very different than the Euclidean space R 1. 2 The upper bounds We start by showing that DP(S) n 2 /4 for every metric space S of size n and doubling dimension 1. The proof uses the following lemma which states that in any isosceles triangle, the base is shorter than the two sides of equal length. Lemma 1 Let S be a finite metric space of doubling dimension 1, and let p, q, and r be three pairwise distinct points in S such that pq = pr. Then qr < pq. Proof. The proof is by contradiction. Let p, q, and r be a smallest triple of points for which the lemma does not hold. That is, let l be the smallest real number such that there exist three points p, q, and r in S such that l = pq = pr and qr l. Let B be the ball with center p and radius l. Observe that B contains the points p, q, and r. Since S has doubling dimension 1, we can cover B by two balls B 1 and B 2 centered at c 1 and c 2, respectively, and having radius l/2. We may assume without loss of generality that p is contained in B 1. Let l = qr. If both q and r are contained in B 2, then l = qr qc 2 + c 2 r l/2 + l/2 = l l. It follows that c 2 q = c 2 r = l/2. Since qr = l l/2, the triple c 2, q, and r forms a smaller counterexample to the lemma, which is a contradiction. Thus, q and r are not both contained in B 2. We may assume without loss of generality that q is contained in B 1. We have l = pq pc 1 + c 1 q l/2 + l/2 = l, which implies that c 1 p = c 1 q = l/2. Since pq = l l/2, the triple c 1, p, and q forms a smaller counterexample to the lemma, which is again a contradiction. 5
6 Theorem 1 Let S be a metric space of doubling dimension 1 and consisting of n points. Then DP(S) n 2 /4. Proof. Let R be the diameter of S, let c be a point in S, and let B be the ball with center c and radius R. Observe that B contains all points of S. Cover B by two balls B 1 and B 2, both having radius R/2, and define S 1 = S B 1 and S 2 = S \ S 1. Assume that S 1 contains two points p and q such that pq = R. Then, denoting the center of B 1 by c, R = pq pc + cq R/2 + R/2 = R and, therefore, cp = cq = R/2, which contradicts Lemma 1. Thus, the diameter of S 1 is less than R. By the same argument, the diameter of S 2 is less than R. If we denote the size of S 1 by m, then it follows that DP(S) is at most m(n m). Since the function m(n m), for 0 m n, is maximized when m = n/2, the proof is complete. The proofs of the upper bounds on the weights of MST (S) and ANN (S) will be by induction on the size of the metric space S. We have to be careful, however, because the doubling dimension d of a subset S of S may be larger than the doubling d dimension of S. The reason is that the value of d is determined by only considering balls that are centered at points of S. To give an example, let n = m 2 and let S be a metric space of size n, in which the distance between any two distinct points is equal to 2. The doubling dimension d of S is equal to d = log n = 2 log m. Partition S into subsets S 1, S 2,..., S m, each consisting of m points. Define S = S {p 1, p 2,..., p m }, where the p i s are new points. For each i with 1 i m, and for each q in S i, we define p i q = qp i = 1. For any other pair of distinct points in S, their distance is defined to be 2. Observe that S is a metric space. We claim that the doubling dimension d of S is equal to d = log(m + 1). To prove this claim, let ρ > 0 be a real number and let B be a ball in S of radius ρ. If ρ 2, then B is covered by m balls of radius ρ/2, centered at the points p 1, p 2,..., p m. If 1 ρ < 2, then B contains at most m + 1 points and, therefore, we can cover B by m + 1 balls of radius ρ/2. Finally, if ρ < 1, then B is a singleton set and we can cover B by one ball of radius ρ/2. Thus, d is indeed equal to log(m + 1). 6
7 For large values of m, the ratio d /d converges (from below) to 2. The next lemma states that this ratio can never be larger than 2. Lemma 2 Let S be a finite metric space of doubling dimension d and let S be a nonempty subset of S. Then the doubling dimension of the metric space S is at most 2d. Proof. Let c be a point of S, let ρ > 0 be a real number, and let B = {p S : cp ρ}. Thus, B is a ball in the metric space S. Consider the corresponding ball B in S, i.e., B = {p S : cp ρ}. Since S has doubling dimension d, we can cover B by balls B 1, B 2,..., B k, all having radius ρ/4, where k 2 2d. For 1 i k, let c i be the center of B i. If c i S, then we define B i to be the ball in S of radius ρ/2 that is centered at c i. If c i S and B i S, then we define B i to be the ball in S of radius ρ/2 that is centered at an arbitrary point of B i S. We claim that the collection of balls {B i : 1 i k, B i S } cover the ball B. To prove this, let p be a point in B. Then p B and, thus, there is an index i such that p B i. If c i S, then p B i. Otherwise, let c i be the center of B i. Since c ip c ic i + c i p ρ/4 + ρ/4 = ρ/2, it follows that p B i. Thus, we have covered the ball B in S by at most 2 2d balls (again, in S ) of radius ρ/2. Theorem 2 Let S be a metric space consisting of n points, let d be the doubling dimension of S, and let R be the radius of S. 1. If d = 1, then wt(mst (S)) 2R log n. 2. If d > 1, then wt(mst (S)) 12Rn 1 1/d. Proof. We start by considering the case when d = 1. We will prove the following claim by induction on the size of V : Let V be a nonempty subset of S and let R V be the radius of V. Then wt(mst (V )) 2R V log V. (1) By taking V = S, this claim will prove the first part of the theorem. 7
8 If V is a singleton set, then (1) obviously holds. Assume that V contains at least two elements and, further, assume that (1) holds for all nonempty subsets of S having less than V elements. Let B be a ball of radius R V that contains V. Since the doubling dimension of S is equal to one, we can cover B by two balls B 1 and B 2, both having radius R V /2. Note that the centers of B 1 and B 2 are not necessarily points of V. Define V 1 = V B 1 and V 2 = V \ V 1. Then both V 1 and V 2 are nonempty (otherwise, the radius of V would be less than R V ). Let T be the spanning tree of V consisting of MST (V 1 ), MST (V 2 ), and one edge joining an arbitrary point of V 1 with an arbitrary point of V 2. Let R V1 and R V2 be the radius of V 1 and V 2, respectively. Then R V1 R V /2 and R V2 R V /2. Since the diameter of V is at most 2R V, and by applying the induction hypothesis, we obtain wt(t ) 2R V1 log V 1 + 2R V2 log V 2 + 2R V R V (log V 1 + log V 2 + 2). Since the function log x is convex for x > 0, and since V 1 + V 2 = V, it follows that wt(t ) R V (log( V /2) + log( V /2) + 2) = 2R V log V. Since wt(mst (V )) wt(t ), we have shown that (1) holds for V. In the rest of the proof, we consider the case when d > 1. We will prove the following claim, again by induction on the size of V : Let V be a nonempty subset of S and let R V be the radius of V. Then wt(mst (V )) 12R V V 1 1/d 12R V. (2) By taking V = S, this claim will prove the second part of the theorem. If V is a singleton set, then wt(mst (V )) = R V = 0 and (2) holds. Assume that V consists of two points p and q. Then wt(mst (V )) = pq 2R V. Thus, (2) holds if we can show that /d 12. This inequality can be rewritten as 7 2 1/d 12. The latter inequality holds, because d log 3. From now on, we assume that V contains at least three elements and (2) holds for all nonempty subsets of S having less than V elements. Let B be a ball of radius R V that contains V. Since the doubling dimension of S is equal to d, we can cover B V by balls B 1, B 2,..., B k, all having radius R V /2, where k 2 d. Observe that, since the radius of V is equal to R V, we 8
9 have k 2. For 1 i k, let V i be the set of points of V that are in the ball B i. In case a point of V is contained in more than one ball, we put it in exactly one subset V i. Thus, the sets V 1, V 2,..., V k are pairwise disjoint and their union is equal to V. We may also assume that none of the sets V i is empty. We next claim that we may assume that k 3. Indeed, if k = 2, then, since V 3, one of the sets V 1 and V 2, say V 2, contains at least two points. We partition V 2 into two disjoint nonempty subsets, which we call V 2 and V 3, and, thus, we can set k = 3. Let i be an index with 1 i k, and let R Vi be the radius of the set V i. Then R Vi R V /2. By the induction hypothesis, we have Since V i 1 1/d 1 0, it follows that wt(mst (V i )) 12R Vi ( Vi 1 1/d 1 ). wt(mst (V i )) 6R V ( Vi 1 1/d 1 ). For each i with 1 i k, choose an arbitrary point in V i, and let T be an arbitrary spanning tree of these k points. Since the diameter of V is at most 2R V, we have wt(t ) 2(k 1)R V 2kR V. Let T be the spanning tree of V consisting of T and MST (V 1 ), MST (V 2 ),..., MST (V k ). Then wt(t ) k ( 6R V Vi 1 1/d 1 ) + 2kR V i=1 k = 6R V V i 1 1/d 4kR V i=1 k 6R V V i 1 1/d 12R V. i=1 Since the function x 1 1/d is convex for x > 0, and since k i=1 V i = V, it follows that k V i 1 1/d i=1 k ( V /k) 1 1/d = k 1/d V 1 1/d 2 V 1 1/d. i=1 9
10 Thus, we have wt(t ) 12R V V 1 1/d 12R V. Since wt(mst (V )) wt(t ), we have shown that (2) holds for V. Theorem 3 Let S be a metric space consisting of n points, where n 2, let d be the doubling dimension of S, and let R be the radius of S. 1. If d = 1, then wt(ann (S)) 4R. 2. If d > 1, then wt(ann (S)) 12Rn 1 1/d. Proof. Since ANN (S) is a subgraph of MST (S), the second claim follows from Theorem 2. Assume that d = 1. The first claim will follow from the following more general claim: For any subset V of S with V 2 and having radius R V, p, NN S (p) 4R V. (3) p V The proof is by induction on the size of V. If V consists of two points a and b, then p V p, NN S(p) 2 ab 4R V, thus (3) holds. Assume that V contains at least three elements and, further, assume that (3) holds for all subsets of S having at least two and less than V elements. Let B be a ball of radius R V that contains V. We cover B by two balls B 1 and B 2, both having radius R V /2, and define V 1 = V B 1 and V 2 = V \ V 1. Observe that both V 1 and V 2 are nonempty. Let R V1 and R V2 be the radius of V 1 and V 2, respectively. Then R V1 R V /2 and R V2 R V /2. If both V 1 and V 2 contain at least two points, then, using the induction hypothesis, p, NN S (p) = p, NN S (p) + p, NN S (p) p V 1 p V 2 p V 4R V1 + 4R V2 4R V. Assume that one of V 1 and V 2, say V 1, consists of one point a. Since the set V has radius R V, we have a, NN S (a) R V. Combining this with the induction hypothesis, we obtain p, NN S (p) = a, NN S (a) + p, NN S (p) R V + 4R V2 4R V. p V 2 p V 10
11 This completes the proof. 3 The lower bounds In this section, we show that the upper bounds in Theorems 1, 2 and 3 are tight. In fact, we give an example of one class of metric spaces that achieve the upper bounds in all these theorems. This class is obtained by repeatedly applying the following transformation: Let m 2 be an integer and assume we are given a metric space S consisting of m points. Let d be the doubling dimension of S, let R be the radius of S, and assume that the diameter of S is equal to R. We define a new metric space S in the following way. For each i with 1 i 2 d, let S i be a copy of the metric space S, and let S be the union of these copies. For any two points p and q in S, if they are in the same copy S i, then the distance pq (in S) is defined to be their distance in S i. Otherwise, we define pq to be 2R. Let n be the size of S, so that n = 2 d m. The following claims can easily be verified. The triangle inequality holds for the set S and, thus, S is a metric space. Both the diameter and the radius of S are equal to R = 2R and ( 2 d DP(S) = 2 ) m 2 = 1 2 (1 12 d ) n 2. (4) Since the diameter of each set S i is equal to R, and since distances between points in S i and points in S j (for j i) are equal to 2R, we have wt(ann (S)) = 2 d wt(ann (S )). (5) By running Kruskal s minimum spanning tree algorithm on S, it follows that wt(mst (S)) = (2 d 1)R + 2 d wt(mst (S )). (6) We finally claim that the doubling dimension of S is equal to d. To prove this, let c be a point in S, let ρ > 0 be a real number, and consider the ball B = {p S : cp ρ}. First assume that ρ < 2R. Then B is completely contained in one of the sets S i. Since the doubling dimension of S i is equal to d, and since distances 11
12 in S i are the same as distances in S, it follows that B can be covered by at most 2 d balls of radius ρ/2. Now assume that ρ 2R. Since the radius of S i is equal to R, there exists a ball B i of radius R that contains S i. The union of these balls B i, for 1 i 2 d, cover the entire set S and, thus, it covers the ball B. Clearly, each such ball has radius at most ρ/2. Thus, also in this case, B can be covered by at most 2 d balls of radius ρ/2. Theorem 4 Let λ 2 be an integer, let d = log λ, let k be a positive integer, and let n = 2 kd. There exists a metric space S of size n and doubling dimension d, such that the following are true: 1. Both the radius and diameter of S are equal to R = 2 k. ( 2. DP(S) = ) /2 d n 2. In particular, if d = 1, then DP(S) = n 2 /4. 3. wt(ann (S)) = 2Rn 1 1/d. In particular, if d = 1, then wt(ann (S)) = 2R. 4. If d = 1, then wt(mst (S)) = R log n. 5. If d > 1, then wt(mst (S)) 2Rn 1 1/d. Proof. Let S be the metric space of size m = 2 d, in which the distance between any two distinct points is equal to 2. The doubling dimension of S is equal to d, and both its radius and diameter are equal to R = 2. By applying the above construction k 1 times, we obtain a metric space S of size n = 2 kd, doubling dimension d, radius R = 2 k and diameter R = 2 k, proving the first claim. The second claim follows from (4). Since wt(ann (S )) = 2m = 2 2 d, it follows from (5) that wt(ann (S)) = 2 2 kd = 2n. Since n = 2 kd and R = 2 k, we have Rn 1 1/d = n. Therefore, we have wt(ann (S)) = 2Rn 1 1/d, proving the third claim. The fourth claim follows from the recurrence in (6). Finally, the fifth claim follows from the third one, because MST (S) contains ANN (S). 4 Concluding remarks We have shown that metric spaces of low doubling dimension are very different than lowdimensional Euclidean spaces. In this final section, we present 12
13 some more consequences of Lemma 1 for metric spaces of doubling dimension 1. Lemma 3 Let S be a finite metric space of doubling dimension 1, and let p, q, and r be three points in S such that (p, q) is an edge in MST (S) and pq pr qr. Then pr 2 pq. Proof. Let l = pq, l = pr, and l = qr. Thus, we have l l l. Let B be the ball with center p and radius l. This ball contains the points p, q, and r. Cover B by balls B 1 and B 2 centered at c 1 and c 2, respectively, and having radius l /2. We may assume without loss of generality that p is contained in B 1. Assume that r is in B 1. Then l = pr pc 1 + c 1 r l /2 + l /2 = l and, therefore, c 1 p = c 1 r = l /2 and pr = l, contradicting Lemma 1. Thus, r is in B 2. If we assume that q is in B 2, then, by a similar argument, it follows that c 2 q = c 2 r = l /2 and qr = l, contradicting Lemma 1. Thus, q is in B 1. We have shown that both p and q are in B 1, and that r is in B 2. Let T = MST (S). By removing the edge (p, q) from T, we obtain two trees T 1 and T 2, where p is a node in T 1 and q is a node in T 2. We may assume without loss of generality that c 1 is a node in T 1. Observe that (c 1, q) is not an edge in T. Let T be the tree obtained from T by replacing the edge (p, q) by the edge (c 1, q). Since T is a minimum spanning tree of S, it follows that l = pq c 1 q. On the other hand, since q is in B 1, we have c 1 q l /2. It follows that l 2l. We now use Lemma 3 to prove the following claims. First, by following any (directed) path in ANN (S), edges decrease in weight by a factor of at least 2. Second, for any point in S, any two incoming edges in ANN (S) differ in weight by a factor of at least 2. Finally, any two edges in MST (S) that share a point differ in length by a factor of at least 2. Lemma 4 Let S be a finite metric space of doubling dimension 1, and let p, q, and r be three pairwise distinct points in S. 1. If q = NN S (p) and p = NN S (r), then pr 2 pq. 13
14 2. If p = NN S (q), p = NN S (r), and pq pr, then pr 2 pq. 3. If both (p, q) and (p, r) are edges in MST (S) and pq pr, then pr 2 pq. Proof. The assumptions in the first and second claims imply that (p, q) is an edge in MST (S) and pq pr qr. Therefore, these two claims follow from Lemma 3. For the third claim, observe that (q, r) is not an edge in MST (S). It follows that pq pr qr and, therefore, the claim also follows from Lemma 3. References [1] P. Assouad. Plongements lipschitziens dans R N. Bulletin de la Société Mathématique de France, 111: , [2] L. Few. The shortest path and the shortest road through n points in a region. Mathematika, 2: , [3] S. HarPeled and M. Mendel. Fast construction of nets in lowdimensional metrics and their applications. SIAM Journal on Computing, 35: , [4] J. Heinonen. Lectures on Analysis on Metric Spaces. SpringerVerlag, Berlin, [5] J. Pach and P. K. Agarwal. Combinatorial Geometry. John Wiley & Sons, New York, [6] M. Smid. The weak gap property in metric spaces of bounded doubling dimension. In Efficient Algorithms, Essays Dedicated to Kurt Mehlhorn on the Occasion of His 60th Birthday, volume 5760 of Lecture Notes in Computer Science, pages , Berlin, SpringerVerlag. [7] W. D. Smith. Studies in Computational Geometry Motivated by Mesh Generation. Ph.D. thesis, Priceton University, Princeton, NJ,
15 [8] J. M. Steele and T. L. Snyder. Worstcase growth rates of some classical problems of combinatorial optimization. SIAM Journal on Computing, 18: ,
arxiv: v1 [cs.cg] 4 Dec 2007
Spanners of Complete kpartite Geometric Graphs Prosenjit Bose Paz Carmi Mathieu Couture Anil Maheshwari Pat Morin Michiel Smid December 17, 013 arxiv:071.0554v1 [cs.cg] 4 Dec 007 Abstract We address the
More informationThe best known algorithm that computes a minimum weight matching is due to Vaidya [6]; its running time is bounded by O(n 5=2 (log n) 4 ) if d = 2, an
A lower bound for approximating the geometric minimum weight matching Gautam Das Michiel Smid y January 6, 1999 Abstract Given a set S of 2n points in R d, a perfect matching of S is a set of n edges such
More informationPlane Geometric Spanners
Plane Geometric Spanners Joachim Gudmundsson, National ICT Australia Ltd, www.gudmundsson.biz Giri Narasimhan, Florida International University, www.cis.fiu.edu/ giri Michiel Smid, Carleton University,
More informationPOINTSETS WITH MANY SIMILAR COPIES OF A PATTERN: WITHOUT A FIXED NUMBER OF COLLINEAR POINTS OR IN PARALLELOGRAMFREE POSITION
POINTSETS WITH MANY SIMILAR COPIES OF A PATTERN: WITHOUT A FIXED NUMBER OF COLLINEAR POINTS OR IN PARALLELOGRAMFREE POSITION BERNARDO M. ÁBREGO AND SILVIA FERNÁNDEZMERCHANT Department of Mathematics
More informationConnectivity, Graph Minors, and Subgraph Multiplicity
Connectivity, Graph Minors, and Subgraph Multiplicity David Eppstein Department of Information and Computer Science University of California, Irvine, CA 92717 January 5, 1993 Abstract It is well known
More informationHOMEWORK 3. e 1 K 1 K 2 SOLUTIONS
HOMEWORK 3 SOLUTIONS (1) Show that for each n N the complete graph K n is a contraction of K n,n. Solution: We describe the process for several small values of n. In this way, we can discern the inductive
More informationGRAPH CONNECTIVITY. 9 Elementary Properties
9 Elementary Properties 30 GRAPH CONNECTIVITY 9 Elementary Properties Definition 9.1: A graph G is said to be connected if for every pair of vertices there is a path joining them. The maximal connected
More informationGraph Theory. Clemens Heuberger and Stephan Wagner. AIMS, January/February 2009
Graph Theory Clemens Heuberger and Stephan Wagner AIMS, January/February 2009 1 Basic Definitions Definition 1.1 A graph is a pair G = (V (G),E(G)) of a set of vertices V (G) and a set of edges E(G), where
More informationThe number of edges in a bipartite graph of given radius
The number of edges in a bipartite graph of given radius P. Dankelmann, Henda C. Swart, P. van den Berg University of KwaZuluNatal, Durban, South Africa Abstract Vizing established an upper bound on the
More informationTwoConnected Steiner Networks: Structural Properties
TwoConnected Steiner Networks: Structural Properties Pawel Winter and Martin Zachariasen Dept. of Computer Science, University of Copenhagen Universitetsparken, DK200 Copenhagen Ø, Denmark email: pawel,martinz
More information2.3 Scheduling jobs on identical parallel machines
2.3 Scheduling jobs on identical parallel machines There are jobs to be processed, and there are identical machines (running in parallel) to which each job may be assigned Each job = 1,,, must be processed
More informationMonochromatic cycle partitions of edgecolored graphs
Monochromatic cycle partitions of edgecolored graphs Gábor N. Sárközy Computer Science Department Worcester Polytechnic Institute Worcester, MA, USA 01609 gsarkozy@cs.wpi.edu and Computer and Automation
More informationThe Steiner ratio for points on a triangular lattice
Volume 24 (2), pp. 185 193 http://www.orssa.org.za ORiON ISSN 0529191X 2008 The Steiner ratio for points on a triangular lattice PO de Wet Received: 26 August 2008; Revised: 16 October 2008; Accepted:
More informationAn easy upper bound for Ramsey numbers
An easy upper bound for Ramsey numbers Roland Bacher To cite this version: Roland Bacher. An easy upper bound for Ramsey numbers. IF PREPUB. This paper contains no new results but presents a slight simplification
More informationSTRONG AND WEAK VITALI PROPERTIES
RESEARCH Real Analysis Exchange Vol. (),, pp. 7 16 Manav Das, Department of Mathematics, University of Louisville, Louisville, Kentucky 40292. email: manav@louisville.edu STRONG AND WEAK VITALI PROPERTIES
More informationDelaunay Triangulation
Delaunay Triangulation Steve Oudot slides courtesy of O. Devillers Outline 1. Definition and Examples 2. Applications 3. Basic properties 4. Construction Definition looking for nearest neighbor Classical
More informationPartitioning edgecoloured complete graphs into monochromatic cycles and paths
arxiv:1205.5492v1 [math.co] 24 May 2012 Partitioning edgecoloured complete graphs into monochromatic cycles and paths Alexey Pokrovskiy Departement of Mathematics, London School of Economics and Political
More informationHomework Assignment #5: Solutions
Advanced Algorithms, Fall 2014 Prof. Bernard Moret Homework Assignment #5: Solutions Question 1. Given a set of points {x 1,...,x n } on the number line, we want to find the smallest set of unitlength
More informationAN OPTIMAL ALGORITHM FOR COMPUTING ANGLECONSTRAINED SPANNERS
AN OPTIMAL ALGORITHM FOR COMPUTING ANGLECONSTRAINED SPANNERS Paz Carmi, and Michiel Smid Abstract. Let S be a set of n points in R d and let t > 1 be a real number. A graph G = (S, E) is called a tspanner
More informationarxiv: v1 [math.co] 26 Jan 2016
Vertex Colorings without Rainbow or Monochromatic Subgraphs Wayne Goddard and Honghai Xu Dept of Mathematical Sciences, Clemson University Clemson SC 29634 {goddard,honghax}@clemson.edu arxiv:1601.06920v1
More informationMinimal double dominating sets in trees
Minimal double dominating sets in trees Marcin Krzywkowski marcin.krzywkowski@gmail.com Abstract. We provide an algorithm for listing all minimal double dominating sets of a tree of order n in time O(1.3248
More informationGraph Theory. 1 Defining and representing graphs
Graph Theory 1 Defining and representing graphs A graph is an ordered pair G = (V, E), where V is a finite, nonempty set of objects called vertices, and E is a (possibly empty) set of unordered pairs
More information2. Lecture notes on nonbipartite matching
Massachusetts Institute of Technology Handout 5 8.433: Combinatorial Optimization February 8th, 009 Michel X. Goemans. Lecture notes on nonbipartite matching Given a graph G = (V, E), we are interested
More informationFigure 1: Example Matching: Matching consists of edges in bold.
CS 6505: Computability & Algorithms Lectures 3 & 4, January 20 & 22, 2010 Greedy Algorithms In this lecture we will examine a couple of famous greedy algorithms and then look at matroids, which are a class
More informationCollege of Computer & Information Science Fall 2007 Northeastern University 9 October 2007
College of Computer & Information Science Fall 2007 Northeastern University 9 October 2007 CS G399: Algorithmic Power Tools I Scribe: Laura Poplawski Lecture Outline: Network Design Steiner Forest (primaldual
More informationThe maximum number of tangencies among convex regions with a trianglefree intersection graph
The maximum number of tangencies among convex regions with a trianglefree intersection graph Eyal Ackerman Abstract Let t(c) be the number of tangent pairs among a set C of n Jordan regions in the plane.
More informationLecture 5. Combinatorial Algorithm for maximum cardinality matching (König s Theorem and Hall s Theorem)
Topics in Theoretical Computer Science March 16, 2015 Lecture 5 Lecturer: Ola Svensson Scribes: Mateusz Golebiewski, Maciej Duleba These lecture notes are partly based on http://theory.epfl.ch/courses/topicstcs/lecture10.pdf
More informationExcluding pairs of graphs
Excluding pairs of graphs Maria Chudnovsky 1 Columbia University, New York, NY 10027, USA Alex Scott Mathematical Institute, University of Oxford, 2429 St Giles, Oxford OX1 3LB, UK Paul Seymour 2 Princeton
More informationLecture 9 and 10: Iterative rounding II
Approximation Algorithms and Hardness of Approximation March 19, 013 Lecture 9 and 10: Iterative rounding II Lecturer: Ola Svensson Scribes: Olivier Blanvillain In the last lecture we saw a framework for
More informationTHE 0/1BORSUK CONJECTURE IS GENERICALLY TRUE FOR EACH FIXED DIAMETER
THE 0/1BORSUK CONJECTURE IS GENERICALLY TRUE FOR EACH FIXED DIAMETER JONATHAN P. MCCAMMOND AND GÜNTER ZIEGLER Abstract. In 1933 Karol Borsuk asked whether every compact subset of R d can be decomposed
More informationUniversity of New Mexico Department of Computer Science. Final Examination. CS 362 Data Structures and Algorithms Spring, 2008
University of New Mexico Department of Computer Science Final Examination CS 362 Data Structures and Algorithms Spring, 2008 Name: Email: Print your name and email, neatly in the space provided above;
More informationL(3, 2, 1)Labeling of Simple Graphs
L(3,, 1)Labeling of Simple Graphs Jean Clipperton Simpson College, IA jean.clipperton@simpson.edu Jessica Gehrtz Concordia CollegeMoorhead, MN jrgehrtz@cord.edu Zsuzsanna Szaniszlo Valparaiso University,
More informationGraph Theory and Metric
Chapter 2 Graph Theory and Metric Dimension 2.1 Science and Engineering Multi discipline teams and multi discipline areas are words that now a days seem to be important in the scientific research. Indeed
More informationCS 372: Computational Geometry Lecture 13 Arrangements and Duality
CS 372: Computational Geometry Lecture 13 Arrangements and Duality Antoine Vigneron King Abdullah University of Science and Technology November 28, 2012 Antoine Vigneron (KAUST) CS 372 Lecture 13 November
More informationOn the Number of Minimal 1Steiner Trees
On the Number of Minimal 1Steiner Trees Boris Aronov Marshall Bern David Eppstein Abstract We count the number of nonisomorphic geometric minimum spanning trees formed by adding a single point to an npoint
More informationGraph Theory. Introduction. Distance in Graphs. Trees. Isabela Drămnesc UVT. Computer Science Department, West University of Timişoara, Romania
Graph Theory Introduction. Distance in Graphs. Trees Isabela Drămnesc UVT Computer Science Department, West University of Timişoara, Romania November 2016 Isabela Drămnesc UVT Graph Theory and Combinatorics
More informationMTH 548 Graph Theory Fall 2003
MTH 548 Graph Theory Fall 2003 Lesson 6  Planar graphs Planar graph plane drawing  plane graph face length of a face infinite face Thm 12: Euler. Let G be a plane drawing of a connected planar graph.
More informationCS 372: Computational Geometry Lecture 11 Voronoi Diagrams and Delaunay Triangulations
CS 372: Computational Geometry Lecture 11 Voronoi Diagrams and Delaunay Triangulations Antoine Vigneron King Abdullah University of Science and Technology November 12, 2012 Antoine Vigneron (KAUST) CS
More informationPlanar Tree Transformation: Results and Counterexample
Planar Tree Transformation: Results and Counterexample Selim G Akl, Kamrul Islam, and Henk Meijer School of Computing, Queen s University Kingston, Ontario, Canada K7L 3N6 Abstract We consider the problem
More informationDefinition A tree is a connected, acyclic graph. Alternately, a tree is a graph in which any two vertices are connected by exactly one simple path.
11.5 Trees Examples Definitions Definition A tree is a connected, acyclic graph. Alternately, a tree is a graph in which any two vertices are connected by exactly one simple path. Definition In an undirected
More informationHonours Graph Theory
University of Zimbabwe HMTH215 Graph Theory Honours Graph Theory Author: P. Mafuta Department: Mathematics April 6, 2016 Chapter 1 Introduction: Basic Graph Theory This course serves to answer many questions
More informationMathematics 1. Part I: Graph Theory. Exercises and problems
Bachelor Degree in Informatics Engineering Barcelona School of Informatics Mathematics 1 Part I: Graph Theory Exercises and problems February 2015 Departament de Matemàtica Aplicada 2 Universitat Politècnica
More informationLecture 7,8 (Sep 22 & 24, 2015):Facility Location, KCenter
,8 CMPUT 675: Approximation Algorithms Fall 2015 Lecture 7,8 (Sep 22 & 24, 2015):Facility Location, KCenter Lecturer: Mohammad R. Salavatipour Scribe: Based on older notes 7.1 Uncapacitated Facility Location
More informationProblem Set 1 Solutions
Design and Analysis of Algorithms February 15, 2015 Massachusetts Institute of Technology 6.046J/18.410J Profs. Erik Demaine, Srini Devadas, and Nancy Lynch Problem Set 1 Solutions Problem Set 1 Solutions
More informationOja Centers and Centers of Gravity
Oja Centers and Centers of Gravity Dan Chen School of Computer Science, Carleton University Olivier Devillers INRIA Sophia Antipolis Méditerranée, France. John Iacono Department of Computer and Information
More informationNew techniques for exact and approximate dynamic closestpoint problems
New techniques for exact and approximate dynamic closestpoint problems Sanjiv Kapoor Michiel Smid August 22, 1995 Abstract Let S be a set of n points in IR D. It is shown that a range tree can be used
More informationMa/CS 6b Class 15: Ramsey Theory 2
Ma/CS 6b Class 15: Ramsey Theory 2 Frank P. Ramsey By Adam Sheffer Recall: Ramsey Numbers r, p 1,, p k positive integers. The Ramsey number R p 1,, p k ; r is the minimum integer N such that every coloring
More informationBACKBONE COLORINGS ALONG STARS AND MATCHINGS IN SPLIT GRAPHS: THEIR SPAN IS CLOSE TO THE CHROMATIC NUMBER
Discussiones Mathematicae Graph Theory 9 (009 ) 14 16 BACKBONE COLORINGS ALONG STARS AND MATCHINGS IN SPLIT GRAPHS: THEIR SPAN IS CLOSE TO THE CHROMATIC NUMBER Hajo Broersma 1, Bert Marchal Daniel Paulusma
More informationHomework MA 725 Spring, 2012 C. Huneke SELECTED ANSWERS
Homework MA 725 Spring, 2012 C. Huneke SELECTED ANSWERS 1.1.25 Prove that the Petersen graph has no cycle of length 7. Solution: There are 10 vertices in the Petersen graph G. Assume there is a cycle C
More informationChapter 11. Approximation Algorithms. Slides by Kevin Wayne PearsonAddison Wesley. All rights reserved.
Chapter 11 Approximation Algorithms Slides by Kevin Wayne. Copyright @ 2005 PearsonAddison Wesley. All rights reserved. 1 Approximation Algorithms Q. Suppose I need to solve an NPhard problem. What should
More informationCOMBINATORIAL PROPERTIES OF THE HIGMANSIMS GRAPH. 1. Introduction
COMBINATORIAL PROPERTIES OF THE HIGMANSIMS GRAPH ZACHARY ABEL 1. Introduction In this survey we discuss properties of the HigmanSims graph, which has 100 vertices, 1100 edges, and is 22 regular. In fact
More informationChapter 5. Connectivity. 5.1 Introduction
Chapter 5 Connectivity 5.1 Introduction The (strong) connectivity corresponds to the fact that a (directed) (u, v)path exists for any pair of vertices u and v. However, imagine that the graphs models
More informationOn the minimal length of the longest trail in a fixed edgedensity graph
Cent Eur J Math 11(10 013 18311837 DOI: 10478/s11533013085x Central European Journal of Mathematics On the minimal length of the longest trail in a fixed edgedensity graph Research Article Vajk Szécsi
More informationMath 443/543 Graph Theory Notes 4: Connector Problems
Math 443/543 Graph Theory Notes 4: Connector Problems David Glickenstein September 19, 2012 1 Trees and the Minimal Connector Problem Here is the problem: Suppose we have a collection of cities which we
More informationThe Longest Cycle of a Graph with a Large Minimal Degree
The Longest Cycle of a Graph with a Large Minimal Degree Noga Alon MATHEMATICS RESEARCH CENTER, AT&T BELL LABORATORIES, MURRAY HILL, NJ 07974 ABSTRACT We show that every graph G on n vertices with minimal
More informationPriority Algorithms. Sashka Davis University of California, San Diego June 2, 2003
Priority Algorithms Sashka Davis University of California, San Diego sdavis@cs.ucsd.edu June 2, 2003 Abstract This paper reviews the recent development of the formal framework of priority algorithms for
More informationConflictfree colorings of graphs
Conflictfree colorings of graphs János Pach EPFL, Lausanne and City College, New York Gábor Tardos Simon Fraser University, Burnaby... Abstract 1 Introduction Let H be a hypergraph with vertex set V (H)
More informationAlgebraic Topology. The Fundamental Group
Algebraic Topology The Fundamental Group Homotopy Given two maps f 1, f 2 : X Y, a homotopy from f 1 to f 2 is a map F : [0, 1] X Y such that F (0, x) = f 1 (x) and F (1, x) = f 2 (x). If there is a homotopy
More informationNeighborhood Unions and Independent Cycles
Neighborhood Unions and Independent Cycles Ronald J. Gould Kazuhide Hirohata Paul Horn Abstract In this paper, we investigate sufficient conditions on the neighborhood of independent vertices which imply
More information1 The Minimum Spanning Tree (MST) Problem
IEOR269: Lecture 3 1 Integer Programming and Combinatorial Optimization Lecture #3 IEOR 269 Feb 5, 2010 Lecturer: Dorit Hochbaum Scribe: Dogan Bilguvar Handouts: Game of Fiver and Fire Station Problems
More informationAll trees contain a large induced subgraph having all degrees 1 (mod k)
All trees contain a large induced subgraph having all degrees 1 (mod k) David M. Berman, A.J. Radcliffe, A.D. Scott, Hong Wang, and Larry Wargo *Department of Mathematics University of New Orleans New
More informationCounting independent sets of a fixed size in graphs with a given minimum degree
Counting independent sets of a fixed size in graphs with a given minimum degree John Engbers David Galvin February 19, 014 Abstract Galvin showed that for all fixed δ and sufficiently large n, the nvertex
More informationThe Pigeonhole Principle
The Pigeonhole Principle 1 Pigeonhole Principle: Simple form Theorem 1.1. If n + 1 objects are put into n boxes, then at least one box contains two or more objects. Proof. Trivial. Example 1.1. Among 13
More informationBipartite Graphs and Problem Solving
Bipartite Graphs and Problem Solving Jimmy Salvatore University of Chicago August 8, 2007 Abstract This paper will begin with a brief introduction to the theory of graphs and will focus primarily on the
More informationINTRODUCTION TO RAMSEY THEORY. Kristina Buschur
INTRODUCTION TO RAMSEY THEORY Kristina Buschur 1. Introduction In general, Ramsey theory deals with the guaranteed occurrence of specific structures in some part of a large arbitrary structure which has
More informationVoronoi diagrams and Delaunay triangulations
Voronoi diagrams and Delaunay triangulations Lecture 9, CS 4235 12 march 2004 Antoine Vigneron antoine@comp.nus.edu.sg National University of Singapore NUS CS4235 Lecture 9: Voronoi diagrams and Delaunay
More informationCourse 421: Algebraic Topology Section 5: Simplicial Complexes
Course 421: Algebraic Topology Section 5: Simplicial Complexes David R. Wilkins Copyright c David R. Wilkins 1988 2008 Contents 5 Simplicial Complexes 58 5.1 Geometrical Independence....................
More informationFinding the Shortest Path Between Two Points in a Simple Polygon By Applying a Rubberband Algorithm
Finding the Shortest Path Between Two Points in a Simple Polygon By Applying a Rubberband Algorithm Fajie Li and Reinhard Klette Computer Science Department, The University of Auckland Auckland, New Zealand
More informationWellSeparated Pair Decomposition for the Unitdisk Graph Metric and its Applications
WellSeparated Pair Decomposition for the Unitdisk Graph Metric and its Applications Jie Gao Department of Computer Science Stanford University Joint work with Li Zhang Systems Research Center HewlettPackard
More informationTopic: PrimalDual Method: Steiner Forest and Facility Location Date: 3/8/2007
CS880: Approximations Algorithms Scribe: Chi Man Liu Lecturer: Shuchi Chawla Topic: PrimalDual Method: Steiner Forest and Facility Location Date: 3/8/2007 Last time we discussed how LPduality can be
More informationCourse Notes for CS336: Graph Theory
Course Notes for CS336: Graph Theory Jayadev Misra The University of Texas at Austin 5/11/01 Contents 1 Introduction 1 1.1 Basics................................. 2 1.2 Elementary theorems.........................
More informationSolutions to Exercises 8
Discrete Mathematics Lent 2009 MA210 Solutions to Exercises 8 (1) Suppose that G is a graph in which every vertex has degree at least k, where k 1, and in which every cycle contains at least 4 vertices.
More informationQuicksort and Randomized Incremental Constructions
Randomized Algorithms, Summer 016 Lecture 3 (7 pages) Quicsort and Randomized Incremental Constructions Instructor: Thomas Kesselheim and Kurt Mehlhorn 1 Quicsort This is Section 5.4 in Mehlhorn/Sanders
More information/ Approximation Algorithms Lecturer: Michael Dinitz Topic: Steiner Tree and TSP Date: 01/29/15 Scribe: Katie Henry
600.469 / 600.669 Approximation Algorithms Lecturer: Michael Dinitz Topic: Steiner Tree and TSP Date: 01/29/15 Scribe: Katie Henry 2.1 Steiner Tree Definition 2.1.1 In the Steiner Tree problem the input
More informationOnline Ramsey Theory for Planar Graphs
Online Ramsey Theory for Planar raphs Šárka Petříčková Department of Mathematics University of West Bohemia Univerzitní 8, 306 14 Plzeň, Czech Republic sarpet@kma.zcu.cz Submitted: Dec 1, 2012; Accepted:
More informationOn the double competition number
On the double competition number Zoltán Füredi Department of Mathematics, University of Illinois, Urbana, IL 61801, USA, and Math. Institute of the Hungarian Academy of Sciences, 1364 Budapest, POB 127
More informationMinimum Spanning Trees
Minimum Spanning Trees Yan Liu LDCSEE West Virginia University, Morgantown, WV {yanliu@csee.wvu.edu} 1 Statement of Problem Let G =< V, E > be a connected graph with realvalued edge weights: w : E R,
More informationCMSC 858F: Algorithmic Game Theory Fall 2010 Frugality & Profit Maximization in Mechanism Design
CMSC 858F: Algorithmic Game Theory Fall 200 Frugality & Profit Maximization in Mechanism Design Instructor: Mohammad T. Hajiaghayi Scribe: Barna Saha September 29th, 200 Overview In this lecture, we start
More informationGraph theory. PoShen Loh. June We begin by collecting some basic facts which can be proved via barehands techniques.
Graph theory PoShen Loh June 011 1 Wellknown results We begin by collecting some basic facts which can be proved via barehands techniques. 1. The sum of all of the degrees is equal to twice the number
More informationFaster Circle Packing with Application to Nonobtuse Triangulation
Faster Circle Packing with Application to Nonobtuse Triangulation David Eppstein Department of Information and Computer Science University of California, Irvine, CA 92717 Tech. Report 9433 August 9, 1994
More informationLecture 12: Steiner Tree Problems
Advanced Approximation Algorithms (CMU 15854B, Spring 008) Lecture 1: Steiner Tree Problems February 1, 008 Lecturer: Anupam Gupta Scribe: Michael Dinitz 1 Finishing up Group Steiner Tree Hardness Last
More informationMinimal forbidden sets for degree sequence characterizations
University of Rhode Island DigitalCommons@URI Mathematics Faculty Publications Mathematics 2015 Minimal forbidden sets for degree sequence characterizations Michael D. Barrus University of Rhode Island,
More information1 Bipartite matching and vertex covers
ORF 523 Lecture 6 Spring 2016, Princeton University Instructor: A.A. Ahmadi Scribe: G. Hall Thursday, February 25, 2016 When in doubt on the accuracy of these notes, please cross check with the instructor
More informationComputing the Diameter of a Point Set
Computing the Diameter of a Point Set Grégoire Malandain and JeanDaniel Boissonnat INRIA, 2004 route des lucioles, BP 93, 06 902 SophiaAntipolis Cedex, France {gregoire.malandain,jeandaniel.boissonnat}@sophia.inria.fr
More informationDefinition 3. We define R(k, m) to be the smallest n such that whenever we 2colour a K n red and blue, we always get either a red K k or a blue K m.
Ramsey Theory We begin with a simple question: How many people do we need to have at a party to ensure that we can find either 3 of them who all know each other, or 3 of them who are all strangers to each
More informationOn the Smallest Minimal Blocking Sets in Projective Space Generating the Whole Space
Beiträge zur Algebra und Geometrie Contributions to Algebra and Geometry Volume 43 (2002), No. 1, 4353. On the Smallest Minimal Blocking Sets in Projective Space Generating the Whole Space Martin Bokler
More informationOn size, radius and minimum degree
Discrete Mathematics and Theoretical Computer Science DMTCS vol. 16:1, 2014, 1 6 On size, radius and minimum degree Simon Mukwembi University of KwaZuluNatal, South Africa. received 9 th Feb. 2012, accepted
More informationIntroduction to Relations
CHAPTER 7 Introduction to Relations 1. Relations and Their Properties 1.1. Definition of a Relation. Definition: A binary relation from a set A to a set B is a subset R A B. If (a, b) R we say a is related
More informationTheorems Matching (independent) edge set A perfect matching edge cover Vertex cover Shadow
Theorems Matching (independent) edge set in agraphis a set ofedges without commonvertices. Aperfect matching is a matching which matches all vertices of the graph. That is, every vertex of the graph isincidentto
More informationHOMEWORK III SOLUTIONS
HOMEWORK III SOLUTIONS CSE 552 RANDOMIED AND APPROIMATION ALGORITHMS. Question. In a Steiner tree of a graph with metric distances, one can always bypass any Steiner vertex of degree 2. If you do this,
More informationApproximating the Minimum Weight Steiner Triangulation
Approximating the Minimum Weight Steiner Triangulation David Eppstein Department of Information and Computer Science University of California, Irvine, CA 92717 November 25, 1992 Abstract We show that the
More informationProperties of Convex Sets: A Glimpse
This is page 62 Printer: Opaque this 3 Properties of Convex Sets: A Glimpse 3.1 Convex Sets Convex sets play a very important role in geometry. In this section we state and prove some of the classics of
More informationOn the Union of Arithmetic Progressions
On the Union of Arithmetic Progressions Shoni Gilboa Rom Pinchasi August, 04 Abstract We show that for any integer n and real ɛ > 0, the union of n arithmetic progressions with pairwise distinct differences,
More informationLengths of geodesics between two points on a Riemannian manifold. Alexander Nabutovsky and Regina Rotman
Lengths of geodesics between two points on a Riemannian manifold. Alexander Nabutovsky and Regina Rotman Abstract. Let x and y be two (not necessarily distinct) points on a closed Riemannian manifold M
More informationOnline Coloring Cointerval Graphs
Online Coloring Cointerval Graphs Hamid ZarrabiZadeh School of Computer Science, University of Waterloo Waterloo, Ontario, Canada, N2L 3G1 hzarrabi@uwaterloo.ca Abstract We study the problem of online
More informationMinimum Spanning Trees. Based on Mahesh Viswanathan s notes
Minimum Spanning Trees Based on Mahesh Viswanathan s notes MST: the problem Input: Connected graph with edge costs Goal: find such that is connected and the total cost of all edges in is the smallest T
More informationGraph theory. PoShen Loh. June 2012
Graph theory PoShen Loh June 2012 At first, graph theory may seem to be an ad hoc subject, and in fact the elementary results have proofs of that nature. The methods recur, however, and the way to learn
More information1 Approximation Algorithms: Vertex Cover
Approximation Algorithms: Vertex Cover. Introduction to Approximation Algorithms There are several optimization problems such as Minimum Spanning Tree (MST), MinCut, Maximum Matching, in which you can
More informationLecture 11: Voronoi diagram, Delaunay triangulation
CPS234 Computational Geometry October 04, 2005 Lecture 11: Voronoi diagram, Delaunay triangulation Lecturer: Pankaj K. Agarwal Scribe: Urmi Majumder 11.1 Introduction Voronoi diagram is the most popular
More information